Warrior General Games

Rnk ..... Off ..... Def .... SoS ........... W-L... Team
1.. ........ 1 ........ 1 ........ 0.410 ........ 10-0.. Deep Throw It
2.. ........ 3 ........ 5 ........ 0.520 ........ 8-2.... Short Bus Window Lickers
3.. ........ 5 ...... 10 ........ 0.460 ........ 8-2.... Atlantic City Surf
4.. ...... 13 ........ 6 ........ 0.400 ........ 9-1.... Georgia Assault
5.. ........ 2 ........ 2 ........ 0.310 ........ 9-1.... Tombstone Outlaws
6.. ...... 22 ........ 9 ........ 0.600 ........ 5-4.... Wood Village Whistle Pigs
7.. ........ 4 ........ 7 ........ 0.360 ........ 9-1.... Fighting Gamecocks
8.. ........ 6 ........ 3 ........ 0.450 ........ 7-2.... Nawlins Pugsimo
9.. ...... 20 ........ 4 ........ 0.510 ........ 7-3.... The Insane Asylum
10 ...... 18 ...... 22 ........ 0.480 ........ 7-3.... Boston Minutemen
11 ...... 17 ...... 25 ........ 0.580 ........ 4-6.... Sacramento VIKINGS Pee Wees
12 ...... 16 ...... 18 ........ 0.470 ........ 6-4.... Deep South Baby Makers
13 ...... 21 ...... 21 ........ 0.560 ........ 5-5.... Traveling Wilburys
14 ...... 10 ........ 8 ........ 0.390 ........ 6-3.... Chicago 85' Bears
15 ...... 19 ...... 11 ........ 0.630 ........ 5-5.... Andromeda Strain
16 ...... 14 ...... 26 ........ 0.480 ........ 5-5.... Dublin Dragons
17 ...... 12 ...... 13 ........ 0.360 ........ 7-3.... Requiem Revolver
18 ........ 9 ...... 16 ........ 0.440 ........ 6-4.... Mississippi Mud Dogs
19 ...... 23 ...... 15 ........ 0.480 ........ 5-5.... Lone Star Mercenaries Mercenary
20 ...... 24 ...... 17 ........ 0.610 ........ 3-7.... Wambulance Academy
21 ...... 11 ...... 12 ........ 0.390 ........ 6-4.... Baltimore Gulls
22 ...... 26 ...... 20 ........ 0.660 ........ 1-9.... East Texas Midnight Riders
23 ........ 8 ...... 19 ........ 0.460 ........ 4-5.... Dharma Initiatives
24 ...... 15 ...... 24 ........ 0.490 ........ 4-6.... Dyslexik Tedd's Monntana Clubb Football
25 ...... 25 ...... 14 ........ 0.450 ........ 4-6.... Hobbittown Hairy Feet
26 ........ 7 ...... 23 ........ 0.380 ........ 4-6.... The Agency
27 ...... 27 ...... 27 ........ 0.680 ........ 0-10.. Marvelous Madmen
28 ...... 28 ...... 30 ........ 0.590 ........ 1-9.... Cameron Copperheads
29 ...... 31 ...... 29 ........ 0.590 ........ 1-9.... Madison Muskies
30 ...... 32 ...... 31 ........ 0.570 ........ 1-9.... He Hate Me !!!
31 ...... 30 ...... 32 ........ 0.580 ........ 0-10.. Boston Patriots
32 ...... 29 ...... 28 ........ 0.460 ........ 1-9.... Bye Week


NEXT WEEK'S MATCHUPS
Exp Score  Matchup
  68 -     6    Tombstone Outlaws def. Boston Minutemen
  39 -   15    Atlantic City Surf def. Lone Star Mercenaries Mercenary
  23 -   13    Nawlins Pugsimo def. Chicago 85' Bears
  59 -   38    Requiem Revolver def. The Agency
  24 -   17    Baltimore Gulls def. Andromeda Strain
114 -     3    Wood Village Whistle Pigs def. Bye Week
  40 -   12    Dharma Initiatives def. East Texas Midnight Riders
  26 -   17    Boston Patriots def. He Hate Me !!!
  78 -     6    Deep Throw It def. Dyslexik Tedd's Monntana Clubb Football
  50 -   12    Fighting Gamecocks def. Traveling Wilburys
  64 -   18    Georgia Assault def. Dublin Dragons
  41 -     7    Short Bus Window Lickers def. Hobbittown Hairy Feet
  37 -   10    The Insane Asylum def. Sacramento VIKINGS Pee Wees
  39 -   21    Mississippi Mud Dogs def. Wambulance Academy
183 -     7    Deep South Baby Makers def. Cameron Copperheads
  48 -     9    Marvelous Madmen def. Madison Muskies


FANTASY MATCHUP OF TOP TEAM IN EACH CONFERENCE
Exp Score  Matchup
  12 -   10    Deep Throw It def. Tombstone Outlaws

Gulls still at 21st?

Strain opponents combined record thus far minus a lone gut job - 55-23-1.

Originally posted by middawg
Strain opponents combined record thus far minus a lone gut job - 55-23-1.

The Gulls will be a good test for you guys...

I've decided the script usually generates scores that are too high (looking at WL and this league). So I changed the algorithm a little bit to favor a good defense over a good offense. These are test predictions (have not updated the script at userscripts.org yet):

NEXT WEEK'S MATCHUPS
Exp Score  Matchup
  42 -     4    Tombstone Outlaws def. Boston Minutemen
  25 -   10    Atlantic City Surf def. Lone Star Mercenaries Mercenary
  14 -     8    Nawlins Pugsimo def. Chicago 85' Bears
  31 -   21    Requiem Revolver def. The Agency
  16 -   11    Baltimore Gulls def. Andromeda Strain
  79 -     2    Wood Village Whistle Pigs def. Bye Week
  27 -     8    Dharma Initiatives def. East Texas Midnight Riders
  18 -   12    Boston Patriots def. He Hate Me !!!
  49 -     4    Deep Throw It def. Dyslexik Tedd's Monntana Clubb Football
  32 -     8    Fighting Gamecocks def. Traveling Wilburys
  40 -   11    Georgia Assault def. Dublin Dragons
  26 -     4    Short Bus Window Lickers def. Hobbittown Hairy Feet
  26 -     7    The Insane Asylum def. Sacramento VIKINGS Pee Wees
  26 -   14    Mississippi Mud Dogs def. Wambulance Academy
126 -     5    Deep South Baby Makers def. Cameron Copperheads
  33 -     6    Marvelous Madmen def. Madison Muskies


FANTASY MATCHUP
Exp Score  Matchup
    7 -     5    Deep Throw It def. Tombstone Outlaws

Obviously the script doesn't look at "football scores", so I'm guessing Tombstone would probably not score exactly 5. Similar with all those 4's up there.

And as someone said in the other thread, this script doesn't take into account levels, CE or effective level; it strictly looks at the scores of previous games in the league. So obviously mid-season gut jobs and significant boosting differences in the middle of a season will have a big impact.

The script doesn't bother trying to predict playoff games at all, for a variety of reasons.

Interesting stuff, guys. And I'll take either score posted above for our game against Lone Star.

I wonder if it's possible to use the new Average Player Value off a team's roster to help in the calculations as it makes a decent sized difference if one team is fully boosted and the other isn't, no matter what their previous games were.

Originally posted by Gerr
I wonder if it's possible to use the new Average Player Value off a team's roster to help in the calculations as it makes a decent sized difference if one team is fully boosted and the other isn't, no matter what their previous games were.
The problem with that is that we don't know when they boosted. If they've been boosted for a while, their record reflects how they perform with boosted players. If they boosted recently, we can expect improvements in their performance from here forward. But GLB (and therefore the scripts that are only pulling information from GLB) doesn't track when boosts were made, so we don't really know at the point that the rankings are generated from the script.

If you really wanted to track that sort of thing, you could make some manual records of the eff level of every team in the league. When a team boosts, you could bump them up a couple of spots in the rankings for a few days, until the impact of that boosting has been accounted for by actual game performance.

And if you want to weigh the last couple of games more heavily (useful to more accurately rank a team that boosted within the past few games), the script allows you to weigh recent games more heavily than older games.

I wasn't talking about tracking the boosting, but instead when the script runs and determines a score, it should factor in the difference in Average Player Value. If they are the same, the script would run as-is. If there is a big difference, then the score needs to be adjusted to show that.

For example, week 11's prediction for my game against Pugsimo...

14 - 8 Nawlins Pugsimo def. Chicago 85' Bears

The current script presumes we are at the same level. It's no secret that my Bears team is fully boosted with a Average Player Value of 259 and the Pugsimo is not boosted with a Average Player Value at 232. That's a pretty big difference that I think the script some how should take that into account.

Granted as players level and boost, the difference will change daily, so just use the current value at the time the script is run.

How exactly the difference in player value is used is up to you, but I do think it should be a factor, if possible.

woot marvelous madmen!

Originally posted by Gerr
I wasn't talking about tracking the boosting, but instead when the script runs and determines a score, it should factor in the difference in Average Player Value. If they are the same, the script would run as-is. If there is a big difference, then the score needs to be adjusted to show that.

For example, week 11's prediction for my game against Pugsimo...

14 - 8 Nawlins Pugsimo def. Chicago 85' Bears

The current script presumes we are at the same level. It's no secret that my Bears team is fully boosted with a Average Player Value of 259 and the Pugsimo is not boosted with a Average Player Value at 232. That's a pretty big difference that I think the script some how should take that into account.

Granted as players level and boost, the difference will change daily, so just use the current value at the time the script is run.

How exactly the difference in player value is used is up to you, but I do think it should be a factor, if possible.

If you include team talent in the rankings and script then it should give you accurate predictions for those who have boosted

My excel program takes these factors in when predicting winners, takes strength of scheule, records, points scored and points allowed and boosted teams etc

An ideal script needs to take into consideration the effective levels of the teams playing that week but also the relative effective levels for the previous games whose stats it is using. The problem there is that that information is not available at the time of the current game being simmed. From what I can tell, LE's program can do this though.

And this doesn't even touch the equipment or token boosting issues that the algorithm cannot know ever (although if everyone opened their builds I guess it could).

Anyway, the point is that there will be a point at which the additional work or variables incorporated into the model will exceed the improved accuracy.

How is strength of schedule calculated? Does it affect the rankings? I'm assuming the lower the number, the lower the strength of schedule. I'll just say take us, DSBM (0.470) for example. 10 games: victory over #2 (now 2 loss team), loss to #7 (one loss team), loss to #9, victory over #11, loss to #13 (ironic that we're ranked behind the team we beat Sacramento and ahead of the team that beat us Wilburys), loss to #18, and victory over #20 who just took down #7. 3 other games are victories over #26, #27, and #29. So, that's 2, 7, 9, 11, 13, 18, 20, 26,27,29 for an average of 16.2.

By contrast, the Whistle Pigs (ranked #6 at 5-4 with a win over #11 The Agency as their highest ranked win) have a SOS of 0.600. They've played # 3, 5, 8, 11, 14, 17, 19, 21, 23, and 26 for an average of 14.7. It's a difference of 1.5 spots which sounds like a lot, but it's not really if you look at it. No doubt they've played a tough schedule, but almost 30% tougher than DSBM?

Is there a way to eliminate teams ranked outside the top 20 from SOS consideration? Typically those are CPU or slow build teams that people don't game plan for nor show their good offense. We all play those teams, just at different times. Eliminating them from the SOS factor would provide a truer reading IMO.

Either way, thanks for doing this. It's fun to see them after each game.

AFAIK, SoS is just opponents wins divided by opponents total games.

So for DSBM:

4-6-0...Agency
9-1-0...GC
6-4-0...MMD
0-10-0.Madmen
4-6-0...SVPW
7-3-0...TIA
1-9-0...Madison
3-7-0...Wambulence
5-5-0...Wilburys
8-2-0...SBWL

Total: 47-53-0 (0.47)

The rankings script is HEAVILY affected by SoS (see these rankings for another league I'm in: http://goallineblitz.com/game/forum_thread.pl?thread_id=3804761) but I don't know exactly how.